本地部署就等于合规Local Deployment Means Compliance

模型部署在本地，合规就自动解决了。Deploy the model locally and compliance is automatically solved.

合规关心的是用途、服务对象、数据获取的授权链条、内容标识与文件留存；这些跟着业务走，不跟着算力位置走。本地部署还要另外处理权重与许可证条款、内部数据接入范围、以及谁在用什么模型。要核对的是「服务如何定性、哪些义务仍适用」，而不是「机器是不是自己的」。（虚构示例）输入是「把同一项对外摘要服务从云端换成本地模型」这个决定 → 处理是照旧核三样：服务面向谁、对外发布的生成内容要不要标识、接入的内部数据授权到哪 → 结果判断是只要有一项仍适用，就不能写「本地部署所以没有义务」→ 下一步是把定性结论交给法务确认，并把标识做法写进发布检查。具体义务以条文原文为准。Compliance follows the business, not the location of the compute: intended use, who the service faces, the authorisation chain for the data, content labeling and record keeping. Local deployment adds its own items - weight and license terms, which internal data gets connected, and who is allowed to use which model. What to verify is how the service is classified and which duties still apply, not whether the machine is yours. (Fictional example) Input is the decision to move the same externally facing summarisation service from cloud to a local model -> processing still checks three things: who the service faces, whether published generated content needs labeling, and how far the internal data authorisation reaches -> if any one of them still applies, the conclusion "local deployment, therefore no duties" is wrong -> the next step is to have legal confirm the classification and to write the labeling practice into the release check. Final obligations follow the text of the rule.